Global environmental impacts: criteria and indicator

“Best bet” Land-use Systems

Country reports

Alternatives To Slash-And-Burn In Indonesia

Unique id: IDAZAQWB

Source file: D:\Projects\ASB\ASB Country and Thematic reports\Indonesia PhaseII report\Part II-III.xml

Authors: Thomas P. Tomich, Meine van Noordwijk, Suseno Budidarsono, Andy Gillison, Trikurnianti Kusumanto, Daniel Murdiyarso, Fred Stolle, Ahmad M. Fagi, Iswandi Anas, A.F.S. Budiman, Kenneth Chomitz, Rebecca Elmhirst, Chip Fay, Hubert de Foresta, Dennis Garrity, Danan P. Hadi, Suryo Hardiwinoto, Kurniatun Hairiah, Genevieve Michon, Nu Nu San, Cheryl Palm, Soetjipto Partoharjono, Djuber Pasaribu, Eric Penot, Robert Simanungkalit, Martua Sirait, S.M. Sitompul, F.X. Susilo, David Thomas

--------------------------------------------------------------------------

Land use at the forest margins has an impact on two global environmental concerns: the net emissions of greenhouse gasses (carbon dioxide, methane and nitrous oxide) which are believed to have an impact on global climate change, and the conservation of biodiversity.

The criterion for effects of land use change on net greenhouse gas emissions can be explained by reference to the effects on natural forests. When considered over large enough scales (in space and/or time) the net carbon exchange between vegetation and atmosphere shows a small flux, equal to the export of organic compounds in soil and water into non-terrestrial ecosystems. The current C stocks in forest systems are large relative to these fluxes and the main issue is in the fate of this stock during land cover change. The two other greenhouse gasses of main global interest (methane and nitrous oxide) can show net emissions or absorption, depending on local soil conditions. Wetland sites (swamp forests as well as rice paddies) generally emit methane, while upland forest soils can absorb and oxidize methane. Nitrous oxide is emitted from all soils where mineral nitrogen is present under relatively wet and warm conditions (so including natural forests), but there may be absorption into green vegetation under certain circumstances. Effects of land use change on greenhouse gas emissions can be measured and expressed in units that allow comparison with industrial emissions, and in the end an economic comparison can be made between the costs of reducing emissions in various sectors of society. Hence, it is important to quantify the effect of land use and land use change on these gasses as fluxes (amount of gas molecules per unit land surface area and unit time).

For biodiversity the criterion is the maintenance of global diversity and the role a particular area plays in that respect, but there is no currency equivalent to the one for greenhouse gases -- diversity measures can be expressed per unit area and per unit time, but can not be converted easily to other units of area or extrapolated in time. For example, if two areas both contain 100 different species, the combined area can contain anywhere between 100 and 200 species, depending on the species overlap. The contribution of a particular site to global biodiversity conservation depends largely on the number of unique flora and fauna elements it contains. Although survey data can show what plants and animals are currently present in a given sampling area, the really important question of how many of these species (or other taxonomic units or genes which are taken as the basis of comparison) would survive over a time frame of X years, can not be directly assessed (Rosenzweig, 1995). Dynamics of local extinction and recolonization depend on the landscape mosaic in which land use systems occur, as well as on the means of dispersal of the organisms concerned. As a very first step into such a dynamic analysis, local species richness is often used as an indicator, largely for lack of better measures. Local species richness can not be compared across ecosystems or even between continents, however, and the best we can do is express local species richness for various land use types relative to that of natural forest. We have to realize, however, that this ratio’s can not be added or subtracted, and that their value probably depends on the scale at which measurements were made. For example, previous comparisons of plant diversity in rubber agroforests showed a local species richness of at least half that of a natural forest, for a 40 m line transect. This does not mean, however, that 1 ha of rubber agroforests will contain (let alone conserve) half the species of 1 ha of natural forest; comparisons at the level of Jambi province are even more uncertain, as it may well be that the 50% forest species in the jungle rubber are generalists, occurring throughout the province and the species not present in the jungle rubber are local specialists, with a different diversity/scale relationship. Despite all these caveats, we will present data here comparing biodiversity indices based on higher plants, which indicate the similarity between sample sites in forest and non-forest, based on a new technique of 'plant functional attributes' (Gillison 1998).

We also collected data on belowground biodiversity, as this is an aspect on which little data exist. Parts of the belowground biodiversity may be directly relevant to the farmer, as they effect 'ecological service functions' (mineralization, soil structure maintenance, symbionts, soil-borne diseases and their control).

II.1 Carbon stocks

Lowland tropical rain forests have the highest standing biomass and aboveground carbon stocks of any vegetation in the world, and total C stocks of rain forests are only equaled by the deepest peat soils. Measurements in Jambi (Fig. II.1) indicate that the total carbon stock of natural forests on the peneplain (above a soil depth of 30 cm) can be up to 50 kg m^-2or 500 Mg ha^-1, with roughly 80% in live trees, 10% in dead wood and 10% in the soil. In logged forests (about 10 years after the logging event), live tree biomass is substantially reduced, but there is more C in dead wood and at least as much in the soil. In cassava fields total C stock can be reduced to about 10% of that in the forest, but soil stocks are still similar to those in the forest. (These data have not been corrected for differences in soil texture, however; compare the Corg/Cref ratio's described in chapter III).

Conversion of rain forest to other land uses, regardless of the technique used for conversion, is thus bound to reduce the amount of C stored in terrestrial ecosystems. As the total net release rate of carbon dioxide (CO₂) into the atmosphere from land use change and fossil fuel emissions exceeds the rate at which the ocean surfaces can absorb additional CO₂, atmospheric CO₂ concentrations increase.

Figure II.1 Carbon stocks in a range of land uses in Jambi

In combinationwith other greenhouse gases, CO₂ is held responsible for increasing the ‘greenhouse effect’ of reflecting radiation from the earth, leading to changes in circulation patterns affecting local climate, as well as causing an overall warming of the planet and an ensuing rise in sea levels. Apart from accepting and adjusting to these climate changes, the main mitigation options are to reduce fossil fuel use and slow down or reverse the trend of declining C stocks in terrestrial ecosystems. In all terrestrial ecosystems C sequestration (fixation) and C dissipation (release) are approximately in equilibrium, with the vast majority of carbon dioxide (CO₂) molecules captured by photosynthesis in leaves during the day being respired at night or during decomposition of litter. Only during phases of build-up of biomass (aboveground or in roots) does the C stock of an ecosystem increase. But in all natural ecosystems, phases of decline and rejuvenation follow phases of growth. And in managed ecosystems, harvest procedures arrest accumulation and usually lead to a period of rejuvenation. In evaluating the C stock of land use systems we have to choose a time frame: following CO₂ molecules at a day or seasonal scale is not necessary, as long as annual increments over the typical life span of a system can be predicted.

Averaging the C stock over the life span of a system gives a simple measure of its role in the global C balance, as long as different stages of the system may be expected to occur in roughly proportional areas at any point in time. If we can assign a typical ‘time-averaged Carbon stock (Mg ha^-1)’ to each land use type, we can directly evaluate how ‘land use change’ will lead to net C release or net C sequestration, depending on the sign of the difference of ‘Cstock(after) – Cstock(before)’.This means that an evaluation of the C stock of a land use depends on the context and the types of comparisons made: compared to natural forest all other land use types lead to net C release to the atmosphere, compared to continuous annual crops, all other land uses lead to C sequestration.

Of particular relevance here may be the C stock of shifting cultivation systems. Fig. II.2 shows how the ‘time-averaged C stock’ depends on the length of fallow and the rate of C sequestration per year during the fallow. For very low land use intensities the time-averaged C stock of shifting cultivation may approach that of a natural forest, as the maximum C stock may be the same and the short episode of slash-and-burn and production of food crops may resemble what happens after a mature tree dies, falls and creates a gap. During intensification of shifting cultivation systems, the time-averaged C stock will decrease rapidly (note the logarithmic scale used for the Y axis in the graph). This analysis emphasizes the systems context of forest clearing: if it is done in the context of long-fallow rotations it will decrease the C stock much less than when it is done for (supposedly) permanent food-crop cultivation.

Figure II.2 Time-averaged Carbon stock of shifting cultivation and fallow rotation systems, as a function of the land use intensity R = Tc/ (Tc + Tf) where Tc is length of cropping period (yr), Tf = length of fallow regrowth period (yr) and Ic = annual C accumulation rate during fallow regrowth (Mg ha^-1 yr^-1)

To estimate the time-averaged C stock of the range of land use systems evaluated as ‘alternatives to slash and burn’, we need the following information:

- Is it a rotational system where periodically whole fields are cleared of vegetation to start a new cycle, or is it managed under permanent vegetation cover?

- What is the length of a single rotation cycle?

- What is the rate of C sequestration per year during the various stages of the cycle (e.g. during periods where annual food crops are grown and during periods of fallow regrowth)?

- Does the C stock reach a maximum at which annual C sequestration levels off?

The land use systems chosen for evaluation all are rotational in nature, except for the community managed forest with extraction of non-timber forest products. Commercial logging (officially) consists of logging episodes and periods where the forest can recover. All other land use systems involve field clearing at the start of a new cycle, mostly using slash-and-burn techniques of land clearing. Some of the rubber agroforests may evolve into a stage of gap-level rejuvenation instead of field level clearing, but the form chosen for evaluation of profitability (chapter IV) is a rotational form. (We will come back to the issue of rotational versus permanent agroforests in chapter IV).

The main remaining uncertainty is the annual rate of C sequestration. The measurements of standing C stock in a range of land uses at different ages since land clearing by slash-and-burn can be used to estimate an average rate of C sequestration (Fig. II.3). In the figure three groups of land use are distinguished:

- logged-over forests; we have to make a rather arbitrary decision on the effective age of the natural forest and the line connecting the points of logged forest with natural forest may overestimate C sequestration if logging has done near-permanent damage to part of the system (such a logging ramps and trails, see chapter III),

- natural fallows (secondary forests), agroforests and more intensive tree-crop production systems, which apparently accumulate at a rate of about 2.5 Mg C

ha-1 yr-1

- cassava/Imperata systems where there is a negligible rate of C accumulation with age, presumably because annual fires prevent the build up of C stocks in vegetation.

On the basis of these results time-averaged C stocks were assigned to the land use types chosen for evaluation (Table II.1).

Fig. II.3 Carbon stock in aboveground biomass, surface litter and top 30 cm of the soil, as a function of time since forest clearing (slash-and-burn) or logging (left: whole data set, right: excluding the natural forest plots)

Table II.1 Time-averaged carbon stocks for land uses of the lowland peneplain; three regression lines were used for the calculations (1 for forest, 2 for agroforest and tree-crop plantations, 3 for cassava-imperata)

Land use system	Line	Maximum age (yr)	Time averaged C stock Mg ha^-1
Natural forest	1	120	254
Community-based forest management	1	60	176
Commercial logging	1	40	150
Rubber agroforests	2	40	116
Rubber agroforests with selected planting material	2	30	103
Rubber monoculture	2	25	97
Oil palm monoculture	2	20	91
Upland rice/ bush fallow rotation	2	7	74
Cassava/Imperata rotation	3	3	39

The values given here contain many assumptions. As part of the ASB-Phase 2 activities in Indonesia, efforts were made to use the Century model (Parton et al., 1987, 1994) for typical transitions from forest into other land use patterns. As the best data on such a time series were collected in the Lampung benchmark area for a forest-to-sugarcane conversion series (where isotope discrimination allows us to follow the fate of 'forest' versus 'cane' organic inputs, Hairiah et al., 1995), model efforts focussed on this series for model validation (Sitompul et al., 1996). Modifications were made to the core routines of the Century model to represent fractions similar to the measurable size-density fractions (LUDOX method, Hairiah et al., 1995). The results (Fig. II.4) show that good agreement between measured data and modelled estimates could indeed be obtained.

When the same model was used, however, for data of the KILLSOM/ADDSOM experiment at the BMSF station in the Lampung benchmark area, agreement between measured and modeled was less convincing (Fig.? II.5); the experimental data contain a substantial scatter, indicating micro-variability not accounted for in the model. Simulations for Peltophorum inputs deviated more from measured points, possibly due to the effect of polyphenolic substances not yet accounted for in the Century model. Overall this experiment shows that none of the organic input treatments is able to maintain the soil organic matter level as it was at the start of the experiment, despite total inputs from litter of at least 8 Mg ha^-1. The main reason for this effect may be a lack of soil macrofauna incorporating litter into the soil -- nearly all inputs decompose at the soil surface and probably contribute little to soil C pools. The century model can be modified to include such effects of soil fauna, and this appears to be a priority area if a better prediction of land use effects on soil carbon pools is needed. Better predictions of soil carbon fractions, however, appear to be more relevant for 'sustainability' issues than they are for the total C balance. Changes in total carbon stocks are clearly dominated by changes in aboveground biomass and a better prediction of vegetation development is the key to improved modeling of land use effects.

II.2 Greenhouse gas emissions

Measurements of the net flux of methane and nitrous oxide were made in a wide range of land use systems. Scaling up from point measurements to typical fluxes over the life span of a land use system (similar to the time-averaged C stock) is not yet possible, however. Day/night as well as seasonal rhythms have to be considered to derive annual flux data, which should be combined for the year of forest clearance and slash-and-burn, early re-growth etc.

Table II.2 summarizes the flux data obtained in the wet and dry season for the land uses of our current evaluation. Methane oxidation rates were higher in the dry than in the wet season. The low level of NH₄ and NO₃ in Imperata and cassava might have caused the low N₂O emission from those land-use systems. Data on N-mineralization, therefore, have to be analyzed to explain the difference with nitrification or denitrification pathways. For the current analysis we explored the relationship between net methane flux and soil bulk density, and between nitrous oxide emission and soil mineral N concentration, both modified by water-filled pore space at the time of observation. Both relationships were weak, and may not form sufficient basis for extrapolation between measuring points. A further process-level analysis of causal factors is probably needed before GHG emissions can be linked to models such as the Century model.

Figure II.4 The dynamics of simulated (lines, S) and observed (points, O) for light (L), intermediate (I), heavy (H) and total macro-organic matter (LIH = L + I + H) fractions when lowland rainforest is converted to sugarcane (A), and the relationship between observed and simulated L, I, H & LIH fractions (B) within 0-20 cm depth under sugarcane. LIH = L + I + H

Figure II.5 Modeled and measured fate of soil macro-organic matter fractions as part of a KILLSOM/ADDSOM experiment in Lampung (Hairiah et al., 1996), where Gliricidia litterfall is the main source of inputs; the overall decline is still a consequence of pastconversion from forests and a lack of incorporation of organic inputs into the soil

Data for methane oxidation and nitrous oxide emission can be compared on the basis of their 'net radiative forcing’ (NRF) CO₂ equivalent values (26 and 206, respectively). It is obvious that removing above-ground carbon stock from forested land or tree-based system will have a greater effect on global warming than that caused by soil emissions. For the natural forest and rubber monoculture plots studied the overall effect on net radiative forcing is negative (this means less global warming, as more methane is oxidized than nitrous oxide emitted in NRF equivalents). For the other land uses nitrous oxide emissions will have a bigger impact on the greenhouse properties of the atmosphere than the methane oxidation.

The last two columns in Table II.2 make a tentative comparison between the greenhouse gas fluxes of land uses per se, with the effects of land use conversions based on change in time-averaged carbon stock. When the difference in C stock is allocated to a 25 year time period, and the data are converted to units of mol C m^-2

yr^-1, it becomes clear that changes in C stock will be one to two orders of magnitude larger than the emissions in the land uses on a stable basis. Obviously, the net climate effect for any land use when derived from lowland rainforest is strongly negative (for the first 25 years), while all land uses would have a substantial mitigating effect on climate change if they replace the Imperata/cassava cycle.

II.3 Belowground biodiversity

Data on belowground biodiversity indicators are summarized in Table II.3. For most parameters the differences between data collected in Jambi and those in Lampung were larger than those between different land uses within each of these benchmark areas. This is reflected in the probability values for the two 'main effects' (province and land use) in table II.3; for a number of parameters land use effects in Lampung differed from those in Jambi, reflected in a statistically significant interaction.

Table II.2 Summary of net greenhouse gas emission effects from current land use (methane and nitrous oxide) and land use change (carbon, allocated to a 25 year period)

Land use system	Time averaged C stock, Mg ha^-1	Mean seasonal net methane absorption, mg m^-2 h^-1			Mean seasonal net N2O emission, mg m^-2 h^-1				Net radiative forcing (C equivalents) mol m^-2 yr^-1
		Wet	Dry		Wet		Dry		soil emissions		LU conversion (25 years)
											from forest		from Imperata
Natural forest	254	0.036		0.046		12.9		1.80		-0.03		0	n.a.
Community-based forest management	176	*		*		*		*		*		26	n.a.
Commercial logging	150	0.044		0.050		17.8		3.60		0.06		35	n.a.
Rubber agroforests	116	0.035		*		34.6		2.97		0.71		46	-26
Rubber agroforests with clonal material	103	*		0.029		*		3.06		0.61		50	-22
Rubber monoculture	97	0.009		0.060		6.1		0.43		-0.06		52	-20
Oil palm monoculture	91	*		*		*		*		*		54	-18
Upland rice/ bush fallow rotation	74	*		*		*		*		*		60	-12
Cassava/Imperata rotation	39	0.001		0.018		9.4		*		0.24		72	0

n.a.= not applicable

*= no data

At first sight the effects of land use on belowground biodiversity appear to be much smaller than expected. Estimates of total population size for most microbial or soil macrofauna groups are remarkably similar, although there are indications of shifts between groups. For example, the Imperata grasslands have the highest densities of earthworms and mycorrhizal spores, while the forests have more ants and spiders in litter and soil samples (but not in the pitfall traps). The total number of soil macrofauna groups present in litter+soil samples was reduced in the Cassava+ Imperata samples, but for pitfall samples no difference was found and for mycorrhizal spore diversity the highest values were found for this land use type.

Table II.3 Results of the surveys of indicators belowground biodiversity in five land uses of the lowland peneplain of Sumatra; the statistical model tested for differences between the two provinces (Lampung versus Jambi, confounded with a different sampling date (September versus November)), five land use categories (Forest, Agroforest, Rehabilitation (young tree-based systems), Cassava and Imperata, respectively) and their interaction. For data on soil fauna the model included a term for depth effects (surface litter and three soil layers), which is not reported here

	Prob of F > value found				Means for land use types
	Province		Land use	P * L	P	all	F	A	R	C	I
Total bacterial count (CFU g^-1 of soil, log)	.0001		.057	.0003	L	3.34	3.48	3.41	4.03	2.49	3.32
					J	4.03	4.00	3.84	3.81	4.21	4.50
					J+L		3.80	3.65	3.94	3.18	3.71
Fungi (CFU g^-1 of soil, log)	.0001		.0008	.0001	L	3.21	3.46	3.39	3.41	2.26	3.44
					J	4.28	3.31	4.10	5.05	5.40	5.11
					J+L		3.37	3.78	4.07	3.52	4.00
Respiration (mg CO2-C kg^-1 day^-1, log)	.0001		.0001	.38	L	1.90	2.04	1.95	2.13	1.48	1.89
					J	2.65	2.83	2.70	2.56	2.33	2.54
					J+L		2.53	2.36	2.30	1.82	2.10
P-solubilizers (CFU, g^-1 of soil, log)	.0001		.0323	.038	L	-1.49	-1.10	-1.80	-0.47	-1.46	-2.38
					J	.376	-.063	0.779	0.897	-.446	0.464
					J+L		-.528	-.510	0.076	-1.21	-1.43
Azotobacter (CFU, g^-1 of soil, log)	.0001		.45	.0004	L	-.167	0.183	0.075	-.243	-1.060	0.036
					J	2.13	1.77	1.72	2.79	2.79	2.50
					J+L		1.17	0.98	1.28	0.59	0.91
Azospirillum (CFU, g^-1 of soil, log)	.0001		.070	.33	L	0.70	1.19	0.417	0.819	0.645	0.416
					J	3.37	3.58	3.14	4.22	4.42	2.11
					J+L		2.22	2.18	1.67	2.53	1.02
Spores of mycorrhizal fungi (g^-1 of soil, log)	.0001		.0001	.0001	L	5.15	4.97	4.80	5.18	5.89	4.96
					J	4.33	3.82	3.80	4.16	5.68	5.60
					J+L		4.25	4.24	4.80	5.81	5.17
Number of mycorrhizal fungal species	.0001		.0001	.0001	L	5.68	5.19	5.89	5.93	6.09	5.39
					J	4.72	4.07	4.08	4.39	5.93	6.89
					L+J		4.49	4.85	5.34	6.04	5.80
Active Soil Carbon indicator 1 (Microb population/Corg )	.28	.59		.41	L	17	11	16	30	12	17
					J	21	15	24	18	29	26
					J+L		14	20	25	19	20
Active Soil Carbon indicator 2 (Microb population * Cref / Corg )	.15		.73	.33	L	43	27	41	82	27	41
					J	61	47	65	43	85	79
					J+L		39	55	66	50	54

PITFALL trappings of active surface fauna (number of individuals per pitfall during 2 days)
Ants (log)	.007	.15	.85	L	4.68	4.76	4.40	5.32	4.28	4.66
				J	5.48	5.56	4.71	6.35	5.41	6.06
				J+L		5.04	4.50	5.51	4.48	4.86
Spiders (log)	.002	.1793	.55	L	2.4	2.37	2.36	3.04	2.46	1.90
				J	3.05	3.02	2.56	3.61	3.26	3.31
				J+L		2.60	2.42	3.15	2.61	2.10
Beetles (log)	.0073	.0154	.77	L	2.54	3.64	1.87	2.98	2.14	2.20
				J	3.76	4.57	3.58	3.36	3.38	3.90
				J+L		3.97	3.39	3.05	3.36	2.30
Cockroaches (log)	.0023	.0021	.46	L	.35	-.03	-.33	.4	.97	.64
				J	.99	.73	.07	2.4	2.0	1.1
				J+L		.24	-.21	.76	1.16	.70
Crickets (log)	.0001	.0001	.57	L	1.93	1.02	.93	2.41	2.92	2.26
				J	3.16	2.71	2.24	3.36	4.47	4.63
				J+L		1.63	1.33	2.58	3.20	2.60
Number of groups per sample	.015	.313	.35	L	5.5	5.3	5.3	6.6	5.6	4.8
				J	6.7	6.6	7.0	6.0	8.0	5.5
				J+L		5.8	5.9	6.5	6.0	4.9
LITTER + SOIL macrofauna (the statistical model included a factor for depth not reported here), No. m^-2
Ants (log)	.73	.0020	.384	L	.26	.75	.39	.31	-.04	0
				J	.50	1.22	.20	.79	.31	-.24
				J+L		1.08	.26	.55	.16	-.12
Spiders (log)	.0001	.0025	.213	L	.25	.62	.79	.04	-.09	-.02
				J	-.32	-.14	-.33	-.29	-.51	-.44
				J+L		.09	.01	-.13	-.33	-.23
Earthworms (log)	.0023	.0064	.049	L	-.18	.15	-.36	-.26	-.55	.03
				J	.34	0	.23	.72	.33	.84
				J+L		.04	.06	.23	-.05	.44
Slugs (log)	.64	.176	.076	L	-.08	.17	.11	.17	0	0
				J	.14	.05	.07	.33	.42	0
				J+L		.08	.08	.25	.24	0
Other groups	.54	.040	.683	L	5.28	8.7	7.3	4.6	4.1	2.6
				J	4.01	5.7	4.8	5.4	1.3	1.3
				J+L		6.6	5.5	5.03	2.5	2.0
Number of groups per sample point	.0001	.0025	.223	L	3.33	4.3	4.0	3.3	2.4	2.9
				J	2.75	3.3	2.5	3.3	2.5	2.1
				J+L		3.5	3.0	3.3	2.5	2.5

In a further analysis of the data we only compared the Imperata/cassava land use (IC) with the three others (RAF). In that analysis we found a significant decrease in IC compared to RAF for respiration, P-solubilizers, woodlice (isopods) caught in pitfall traps and ants, spiders, cockroaches, crickets, 'other' and group diversity for the soil macrofauna. A statistically significant increase was found for mycorrhizal spore density and diversity and pitfall catches of cockroaches, slugs and crickets. For parameters such as earthworms an increase in Imperata was off-set by a decrease in cassava.

In the Lampung benchmark area detailed information was obtained on nematode genera (or families) in the five (ICRAF) land uses. Only for the plant-parasitic Meloidogyne nematodes did we find a significant (p < .001) effect of land use, with very high densities in the cassava fields, intermediate ones in the forested fields (RAF) and an absence in the Imperata fallow plots. For the other groups (Rhabditida, Dorylaimida, Criconemoides, Tylenchus, Helicotylenchus, Rotylenchus, Monochus, Hoplolaimus, Scutelonema, Aphelenchus) differences between replicate samples in the same land use were larger than those between land uses as a group, so the null-hypothesis of no land use effect was not rejected.

The number of rhizobia in the soil was estimated using a MPN method (Brockwell et al., 1975) and three legumes (Macroptilium atropurpureum, Pueraria phaseoloides and Glycinesoja) as host plants. Siratro-nodulating bacteria were found in only one location of forest, and mature agroforest, all three locations of young agroforest, two locations of cassava and two location of Imperata grasslands, while kudzu-nodulating bacteria were found in one location of forest, one location of mature agroforest, two locations of young agroforest, none of cassava and Imperata grasslands. There were no wild soybean-nodulating bacteria found in any locations in Lampung. In Jambi siratro-nodulating bacteria were found in two of the four locations of forest, one of the five locations of mature agroforest, one of the two locations of young agroforest, none of the two locations of cassava, and one of two locations of Imperata grassland. Kudzu-nodulating bacteria were found in two of the four locations of forest, more of the five locations of mature agroforest, one of the two locations of young agroforest, none in cassava and Imperata grasslands. Similarly, wild soybean-nodulating bacteria were not found in any locations in Jambi. The results thus indicate that in several locations land use systems are lacking suitable host legumes . Importantly, there were no indications of a relationship between occurrence of symbiotic N2-fixing bacteria and land use system. The occurrence of symbiotic N2-fixing bacteria seems to be influenced by the presence of suitable host legumes in the respective land use systems.

It may be that our conclusion of relatively small effects of land use on soil fauna is colored by the type of parameters measured. It is possible that greater differences would appear if more sensitive parameters were collected, e.g. specific groups of spiders and ants rather than the groups as a whole. Some evidence on much stronger response to land use change was collected as part of the intensive biodiversity survey in Jambi, where termite data were collected and sorted by tropic group (wood versus soil feeders). These (un-replicated) samples showed large differences between forest and agroforests on one hand and the cassava/Imperata plots in the other hand (Swift 1998).

II.4 Aboveground biodiversity

As part of the integrated survey of land use systems in the peneplains, aboveground biodiversity was assessed in terms of the richness of species and plant functional types (‘modi’) in standard-sized sample plots. In the data analysis a single vector ‘V index’ may be defined which gives a clear differentiation between Imperata grasslands as one extreme and natural forest as the other. The vector is composed of a large number of the plot-level measurements (Fig. II.6).

The V index classifies monospecific tree plantations with their associated ‘weeds’ as halfway on the scale between natural forest and Imperata grasslands, close to the vegetation of a logging ramp as part of logged forests. Old rubber agroforests are intermediate between logged and natural parts of natural forest, confirming earlier data on species richness (De Foresta and Michon, 1997). The V-index is based on a number of parameters, including basal area of trees, plant species richness and number of unique combinations (modi) of plant functional attributes (PFA). PFA diversity of rubber agroforests can equal that of natural forests, but the number of botanical plant species per modus is less. The data suggest that the ratio of botanical species and modi may be an informative single indicator of aboveground biodiversity of forests and forest-derived land covers. As may be expected, a good correlation exists between aboveground C stock and such indices of aboveground (plant) biodiversity (Fig. II.7).

Figure II.6 Overall classification of vegetation structure and plant biodiversity ('V' index) for intensive sampling points in Jambi; the V index is the most-discriminating single axis in multidimensional parameter space, which groups 'similar' plots

II.5 Landscape level assessments

Some first steps were made towards landscape level diversity assessments, including diversity among different sample points in the same land use class. The basic question may be phrased as: are all forest sites the same ('if you've seen one forest you've seen them all') or do they contain more internal variation then human-derived land covers, with the Imperata/ cassava system as extreme.

Figure II.7 Comparative relationships between above-ground carbon, plant functional type richness, species richness and species / modi ratios along a gradient of land use types, Jambi, Lowland Sumatra

Figure II.8Ordination (showing the two first principal components) of sample points for all parameters in the integrated survey (abiotic + vegetation + soil) or different subsets of these parameters; the lines indicate the domains for forest sample points as natural background and Imperata + cassava as extremes of human modification, I= Imperata, C=Cassava, R= Rehabilitation (young AF system), A= Agroforest, F= Forest, L= Lampung (open symbols), J= Jambi (closed symbols)

Figure II.8 presents the 31 points for the integrated survey, using different parts of the total data set for defining similarity among sample points. If only the abiotic soil parameters are considered, the area spanned by the forest points more or less coincides with that of the cassava/Imperata system, indicating that basic soil characteristics are probably little changed by forest conversion (upper right in Fig. II.8). The Lampung points (open symbols) fall in a different class than the Jambi points (closed symbols), and this dichotomy is conserved for all other parts of the data set. If the soil biological parameters are added to the abiotic soil descriptors (see lower right quadrant), the Imperata/cassava points stand a bit further out from the forest ones, but there are no simple tests of the statistical significance of such a difference. When the vegetation parameters are combined with abiotic soil descriptors (lower left), the cassava/Imperata points for Lampung are clearly outside the forest points, indicating that this conversion may have increased landscape level diversity. When all parameters are considered (upper left), distances are less pronounced.

The view that part of the 'savanization' (formation of grasslands) of forests can be seen as an increase of landscape level diversity is supported by analyses of large mammals in a landscape historical context. Boomgaard (1997) argued that large mammal populations initially benefited from human presence in forest landscapes.

The transformation of forests into agroforests may initially have added little to landscape level diversity, in the sense that all parameter combinations found in such agroforests are within the domain of natural forests. During this transformation, these agroforests have become a major reservoir for forest flora and fauna in the current landscape where natural forest has become scarce (Jambi) or near absent (Lampung). Current data indicate that old rubber agroforests indeed contain a substantial part of forest diversity. However, more detailed research on fern diversity (H. Beukema, research in progress) shows that the between-plot variation in species composition of natural forests is substantially larger than that for rubber agroforests, even if plot-level diversity is approximately the same. Translating the current plot-level assessments to landscape level statements about global environmental impacts is thus not a trivial exercise, which will need further attention in future assessments.